Real-time inference engine: trained transformer for dynamic prediction

Third article in series demonstrates how a trained Transformer can perform real-time inference without perceptible latency.